I recently ran multiple regression analysis in the WJ III norm data [conflict of interest note - I'm a coauthor of the WJ III] where I used the complete set of WJ III cognitive tests (the original WJ III and WJ III Diagnostic Supplement tests) and WJ III oral language tests to predict the WJ III (a) individual reading tests, (b) reading clusters, (c) individual math tests, and (d) math clusters.
Summary tables of the results are now included in the CHC reading and math summary visual-graphic mindmaps posted previously. As noted in the summary documents, I ran step-wise multiple regression models (with backward stepping) at three different age groups in the WJ III norm sample (ages 6-8; 9-13; 14-19). I specified that the models include five predictor tests. Due to possible predictor-criterion contamination, the WJ III Number Series test was excluded from the predictor (IV) pool in the prediction of the Math Reasoning and Quantitative Concepts test (half of the QC test, which is part of the MR cluster, includes number series type items). Finally, the regression models were run on correlation matrices that were calculated in each age group with age variance removed (age-based standard scores were used). The EM missing data algorithm was invoked during the calculation of the matrices. These matrices where then used for the multiple regression analysis.
Descriptions and explanations of the WJ III tests are available via a link in a prior post (first link under "The Results: Phase I" section of prior post)
The summary tables include the final standardized regression weights as well as the coding of tests that, although not ending up in the final regression models, were often close to entering the regression model at some steps (based on inspection of the partial correlations at each step)--call them bridesmaid tests.
This material is being incorporated into a presentation I'm completing as part of a NASP 2008 workshop. The interpretation of these results, combined with a select review and synthesis of CHC- and non-CHC-organized COG-ACH correlates research, will be presented at that workshop, as well as the eventual posting of select PPT slides from that workshop (watch for announcements at this blog).
Please recognize that these results have NOT been peer-reviewed. The results are being presented "as is" with no interpretation. As noted above, interpretation of this material will be part of the NASP workshop. Additionally, this material will be included in the next update to the WJ III Evolving Web of Knowledge (EWOK), which I plan to update prior to (or immediately after) this workshop.
Enjoy running your fingers through the analysis summaries. I hope the posting of this information stimulates hypothesis generation and discussion by other CHC/WJ III scholars and assessment professionals. I urge interested individuals to make comments on the CHC listserv...as the CHC listserv provides for a more dynamic give-and-take learning experience than available via static blog "comment" posts.
You can go directly past go and download the reports at the following two links.
- WJ III reading achievement multiple regression summary
- WJ III math achievement multiple regression summary
Technorati Tags: psychology, educational psychology, school psychology, neuropsychology, Riverside Publishing, Woodcock-Johnson, WJ III, WJ III NU, normative update, reading, math, Grw, Gq, CHC, CHC theory, multiple regression
Powered by ScribeFire.